The purpose of this study was to investigate approaches evaluating the performance of safety systems in crash tests and by analytical simulations. The study was motivated by the need to consider the adequacy of injury criteria and tolerance levels in FMVSS 208 measuring safety performance of restraint systems and supplements. The study also focused on additional biomechanical criteria and performance measures which may augment FMVSS 208 criteria and alternative ways to evaluate dummy responses rather than by comparison to a tolerance level.Additional analysis was conducted of dummy responses from barrier crash and sled tests to gain further information on the performance of restraint systems. The analysis resulted in a new computer program which determined several motion and velocity criteria from measurements made in crash tests. These data provide new insights on restraint performance including torso angle change, the effectiveness of occupant restraint from velocity buildup -- called the Restraint Quotient -- and forward displacement and rebound of the chest and pelvis. The various approaches are used to discuss Hybrid III responses from 24 barrier crash and 23 Hyge sled tests.Four additional experiments were critically evaluated to validate the new analysis. They involved a lap-shoulder belted Hybrid III female dummy with the lap-belt positioned low on the pelvis for restraint or pre-positioned on the abdomen for submarining. By comparing calculated and film analyzed data, the Restraint Quotient program was found to predict accurately those test results.Numbers relating to FMVSS 208 criteria values were reduced 15%-25% by lap-belt submarining over lap-belt restraint. The addition of an abdominal injury criterion and use of a frangible abdomen helped predict injury risk to the abdomen and an overall higher risk of injury with lap-belt submarining. The use of motion and velocity criteria also identified poorer performance with lap-belt submarining.Several approaches were reviewed to interpret the overall performance and injury risks of safety systems in a crash. The preferred method was found to be injury risk assessment which represents dummy responses as injury risks using Logist probability functions. This enables injury in each body region to be assessed by the maximum risk from applicable biomechanical criteria. Whole body injury is determined by summing the individual risks from each body region.The motion and velocity criteria determined from the Restraint Quotient program were useful in complementing the biomechanical criteria to assess injury. They may be helpful in an overall determination of the crash performance of restraint systems, supplements and enhancements.